Added license information
[kvmfornfv.git] / kernel / init / initramfs.c
1 /*
2  * Many of the syscalls used in this file expect some of the arguments
3  * to be __user pointers not __kernel pointers.  To limit the sparse
4  * noise, turn off sparse checking for this file.
5  */
6 #ifdef __CHECKER__
7 #undef __CHECKER__
8 #warning "Sparse checking disabled for this file"
9 #endif
10
11 #include <linux/init.h>
12 #include <linux/fs.h>
13 #include <linux/slab.h>
14 #include <linux/types.h>
15 #include <linux/fcntl.h>
16 #include <linux/delay.h>
17 #include <linux/string.h>
18 #include <linux/dirent.h>
19 #include <linux/syscalls.h>
20 #include <linux/utime.h>
21
22 static ssize_t __init xwrite(int fd, const char *p, size_t count)
23 {
24         ssize_t out = 0;
25
26         /* sys_write only can write MAX_RW_COUNT aka 2G-4K bytes at most */
27         while (count) {
28                 ssize_t rv = sys_write(fd, p, count);
29
30                 if (rv < 0) {
31                         if (rv == -EINTR || rv == -EAGAIN)
32                                 continue;
33                         return out ? out : rv;
34                 } else if (rv == 0)
35                         break;
36
37                 p += rv;
38                 out += rv;
39                 count -= rv;
40         }
41
42         return out;
43 }
44
45 static __initdata char *message;
46 static void __init error(char *x)
47 {
48         if (!message)
49                 message = x;
50 }
51
52 /* link hash */
53
54 #define N_ALIGN(len) ((((len) + 1) & ~3) + 2)
55
56 static __initdata struct hash {
57         int ino, minor, major;
58         umode_t mode;
59         struct hash *next;
60         char name[N_ALIGN(PATH_MAX)];
61 } *head[32];
62
63 static inline int hash(int major, int minor, int ino)
64 {
65         unsigned long tmp = ino + minor + (major << 3);
66         tmp += tmp >> 5;
67         return tmp & 31;
68 }
69
70 static char __init *find_link(int major, int minor, int ino,
71                               umode_t mode, char *name)
72 {
73         struct hash **p, *q;
74         for (p = head + hash(major, minor, ino); *p; p = &(*p)->next) {
75                 if ((*p)->ino != ino)
76                         continue;
77                 if ((*p)->minor != minor)
78                         continue;
79                 if ((*p)->major != major)
80                         continue;
81                 if (((*p)->mode ^ mode) & S_IFMT)
82                         continue;
83                 return (*p)->name;
84         }
85         q = kmalloc(sizeof(struct hash), GFP_KERNEL);
86         if (!q)
87                 panic("can't allocate link hash entry");
88         q->major = major;
89         q->minor = minor;
90         q->ino = ino;
91         q->mode = mode;
92         strcpy(q->name, name);
93         q->next = NULL;
94         *p = q;
95         return NULL;
96 }
97
98 static void __init free_hash(void)
99 {
100         struct hash **p, *q;
101         for (p = head; p < head + 32; p++) {
102                 while (*p) {
103                         q = *p;
104                         *p = q->next;
105                         kfree(q);
106                 }
107         }
108 }
109
110 static long __init do_utime(char *filename, time_t mtime)
111 {
112         struct timespec t[2];
113
114         t[0].tv_sec = mtime;
115         t[0].tv_nsec = 0;
116         t[1].tv_sec = mtime;
117         t[1].tv_nsec = 0;
118
119         return do_utimes(AT_FDCWD, filename, t, AT_SYMLINK_NOFOLLOW);
120 }
121
122 static __initdata LIST_HEAD(dir_list);
123 struct dir_entry {
124         struct list_head list;
125         char *name;
126         time_t mtime;
127 };
128
129 static void __init dir_add(const char *name, time_t mtime)
130 {
131         struct dir_entry *de = kmalloc(sizeof(struct dir_entry), GFP_KERNEL);
132         if (!de)
133                 panic("can't allocate dir_entry buffer");
134         INIT_LIST_HEAD(&de->list);
135         de->name = kstrdup(name, GFP_KERNEL);
136         de->mtime = mtime;
137         list_add(&de->list, &dir_list);
138 }
139
140 static void __init dir_utime(void)
141 {
142         struct dir_entry *de, *tmp;
143         list_for_each_entry_safe(de, tmp, &dir_list, list) {
144                 list_del(&de->list);
145                 do_utime(de->name, de->mtime);
146                 kfree(de->name);
147                 kfree(de);
148         }
149 }
150
151 static __initdata time_t mtime;
152
153 /* cpio header parsing */
154
155 static __initdata unsigned long ino, major, minor, nlink;
156 static __initdata umode_t mode;
157 static __initdata unsigned long body_len, name_len;
158 static __initdata uid_t uid;
159 static __initdata gid_t gid;
160 static __initdata unsigned rdev;
161
162 static void __init parse_header(char *s)
163 {
164         unsigned long parsed[12];
165         char buf[9];
166         int i;
167
168         buf[8] = '\0';
169         for (i = 0, s += 6; i < 12; i++, s += 8) {
170                 memcpy(buf, s, 8);
171                 parsed[i] = simple_strtoul(buf, NULL, 16);
172         }
173         ino = parsed[0];
174         mode = parsed[1];
175         uid = parsed[2];
176         gid = parsed[3];
177         nlink = parsed[4];
178         mtime = parsed[5];
179         body_len = parsed[6];
180         major = parsed[7];
181         minor = parsed[8];
182         rdev = new_encode_dev(MKDEV(parsed[9], parsed[10]));
183         name_len = parsed[11];
184 }
185
186 /* FSM */
187
188 static __initdata enum state {
189         Start,
190         Collect,
191         GotHeader,
192         SkipIt,
193         GotName,
194         CopyFile,
195         GotSymlink,
196         Reset
197 } state, next_state;
198
199 static __initdata char *victim;
200 static unsigned long byte_count __initdata;
201 static __initdata loff_t this_header, next_header;
202
203 static inline void __init eat(unsigned n)
204 {
205         victim += n;
206         this_header += n;
207         byte_count -= n;
208 }
209
210 static __initdata char *vcollected;
211 static __initdata char *collected;
212 static long remains __initdata;
213 static __initdata char *collect;
214
215 static void __init read_into(char *buf, unsigned size, enum state next)
216 {
217         if (byte_count >= size) {
218                 collected = victim;
219                 eat(size);
220                 state = next;
221         } else {
222                 collect = collected = buf;
223                 remains = size;
224                 next_state = next;
225                 state = Collect;
226         }
227 }
228
229 static __initdata char *header_buf, *symlink_buf, *name_buf;
230
231 static int __init do_start(void)
232 {
233         read_into(header_buf, 110, GotHeader);
234         return 0;
235 }
236
237 static int __init do_collect(void)
238 {
239         unsigned long n = remains;
240         if (byte_count < n)
241                 n = byte_count;
242         memcpy(collect, victim, n);
243         eat(n);
244         collect += n;
245         if ((remains -= n) != 0)
246                 return 1;
247         state = next_state;
248         return 0;
249 }
250
251 static int __init do_header(void)
252 {
253         if (memcmp(collected, "070707", 6)==0) {
254                 error("incorrect cpio method used: use -H newc option");
255                 return 1;
256         }
257         if (memcmp(collected, "070701", 6)) {
258                 error("no cpio magic");
259                 return 1;
260         }
261         parse_header(collected);
262         next_header = this_header + N_ALIGN(name_len) + body_len;
263         next_header = (next_header + 3) & ~3;
264         state = SkipIt;
265         if (name_len <= 0 || name_len > PATH_MAX)
266                 return 0;
267         if (S_ISLNK(mode)) {
268                 if (body_len > PATH_MAX)
269                         return 0;
270                 collect = collected = symlink_buf;
271                 remains = N_ALIGN(name_len) + body_len;
272                 next_state = GotSymlink;
273                 state = Collect;
274                 return 0;
275         }
276         if (S_ISREG(mode) || !body_len)
277                 read_into(name_buf, N_ALIGN(name_len), GotName);
278         return 0;
279 }
280
281 static int __init do_skip(void)
282 {
283         if (this_header + byte_count < next_header) {
284                 eat(byte_count);
285                 return 1;
286         } else {
287                 eat(next_header - this_header);
288                 state = next_state;
289                 return 0;
290         }
291 }
292
293 static int __init do_reset(void)
294 {
295         while (byte_count && *victim == '\0')
296                 eat(1);
297         if (byte_count && (this_header & 3))
298                 error("broken padding");
299         return 1;
300 }
301
302 static int __init maybe_link(void)
303 {
304         if (nlink >= 2) {
305                 char *old = find_link(major, minor, ino, mode, collected);
306                 if (old)
307                         return (sys_link(old, collected) < 0) ? -1 : 1;
308         }
309         return 0;
310 }
311
312 static void __init clean_path(char *path, umode_t fmode)
313 {
314         struct stat st;
315
316         if (!sys_newlstat(path, &st) && (st.st_mode ^ fmode) & S_IFMT) {
317                 if (S_ISDIR(st.st_mode))
318                         sys_rmdir(path);
319                 else
320                         sys_unlink(path);
321         }
322 }
323
324 static __initdata int wfd;
325
326 static int __init do_name(void)
327 {
328         state = SkipIt;
329         next_state = Reset;
330         if (strcmp(collected, "TRAILER!!!") == 0) {
331                 free_hash();
332                 return 0;
333         }
334         clean_path(collected, mode);
335         if (S_ISREG(mode)) {
336                 int ml = maybe_link();
337                 if (ml >= 0) {
338                         int openflags = O_WRONLY|O_CREAT;
339                         if (ml != 1)
340                                 openflags |= O_TRUNC;
341                         wfd = sys_open(collected, openflags, mode);
342
343                         if (wfd >= 0) {
344                                 sys_fchown(wfd, uid, gid);
345                                 sys_fchmod(wfd, mode);
346                                 if (body_len)
347                                         sys_ftruncate(wfd, body_len);
348                                 vcollected = kstrdup(collected, GFP_KERNEL);
349                                 state = CopyFile;
350                         }
351                 }
352         } else if (S_ISDIR(mode)) {
353                 sys_mkdir(collected, mode);
354                 sys_chown(collected, uid, gid);
355                 sys_chmod(collected, mode);
356                 dir_add(collected, mtime);
357         } else if (S_ISBLK(mode) || S_ISCHR(mode) ||
358                    S_ISFIFO(mode) || S_ISSOCK(mode)) {
359                 if (maybe_link() == 0) {
360                         sys_mknod(collected, mode, rdev);
361                         sys_chown(collected, uid, gid);
362                         sys_chmod(collected, mode);
363                         do_utime(collected, mtime);
364                 }
365         }
366         return 0;
367 }
368
369 static int __init do_copy(void)
370 {
371         if (byte_count >= body_len) {
372                 if (xwrite(wfd, victim, body_len) != body_len)
373                         error("write error");
374                 sys_close(wfd);
375                 do_utime(vcollected, mtime);
376                 kfree(vcollected);
377                 eat(body_len);
378                 state = SkipIt;
379                 return 0;
380         } else {
381                 if (xwrite(wfd, victim, byte_count) != byte_count)
382                         error("write error");
383                 body_len -= byte_count;
384                 eat(byte_count);
385                 return 1;
386         }
387 }
388
389 static int __init do_symlink(void)
390 {
391         collected[N_ALIGN(name_len) + body_len] = '\0';
392         clean_path(collected, 0);
393         sys_symlink(collected + N_ALIGN(name_len), collected);
394         sys_lchown(collected, uid, gid);
395         do_utime(collected, mtime);
396         state = SkipIt;
397         next_state = Reset;
398         return 0;
399 }
400
401 static __initdata int (*actions[])(void) = {
402         [Start]         = do_start,
403         [Collect]       = do_collect,
404         [GotHeader]     = do_header,
405         [SkipIt]        = do_skip,
406         [GotName]       = do_name,
407         [CopyFile]      = do_copy,
408         [GotSymlink]    = do_symlink,
409         [Reset]         = do_reset,
410 };
411
412 static long __init write_buffer(char *buf, unsigned long len)
413 {
414         byte_count = len;
415         victim = buf;
416
417         while (!actions[state]())
418                 ;
419         return len - byte_count;
420 }
421
422 static long __init flush_buffer(void *bufv, unsigned long len)
423 {
424         char *buf = (char *) bufv;
425         long written;
426         long origLen = len;
427         if (message)
428                 return -1;
429         while ((written = write_buffer(buf, len)) < len && !message) {
430                 char c = buf[written];
431                 if (c == '0') {
432                         buf += written;
433                         len -= written;
434                         state = Start;
435                 } else if (c == 0) {
436                         buf += written;
437                         len -= written;
438                         state = Reset;
439                 } else
440                         error("junk in compressed archive");
441         }
442         return origLen;
443 }
444
445 static unsigned long my_inptr; /* index of next byte to be processed in inbuf */
446
447 #include <linux/decompress/generic.h>
448
449 static char * __init unpack_to_rootfs(char *buf, unsigned long len)
450 {
451         long written;
452         decompress_fn decompress;
453         const char *compress_name;
454         static __initdata char msg_buf[64];
455
456         header_buf = kmalloc(110, GFP_KERNEL);
457         symlink_buf = kmalloc(PATH_MAX + N_ALIGN(PATH_MAX) + 1, GFP_KERNEL);
458         name_buf = kmalloc(N_ALIGN(PATH_MAX), GFP_KERNEL);
459
460         if (!header_buf || !symlink_buf || !name_buf)
461                 panic("can't allocate buffers");
462
463         state = Start;
464         this_header = 0;
465         message = NULL;
466         while (!message && len) {
467                 loff_t saved_offset = this_header;
468                 if (*buf == '0' && !(this_header & 3)) {
469                         state = Start;
470                         written = write_buffer(buf, len);
471                         buf += written;
472                         len -= written;
473                         continue;
474                 }
475                 if (!*buf) {
476                         buf++;
477                         len--;
478                         this_header++;
479                         continue;
480                 }
481                 this_header = 0;
482                 decompress = decompress_method(buf, len, &compress_name);
483                 pr_debug("Detected %s compressed data\n", compress_name);
484                 if (decompress) {
485                         int res = decompress(buf, len, NULL, flush_buffer, NULL,
486                                    &my_inptr, error);
487                         if (res)
488                                 error("decompressor failed");
489                 } else if (compress_name) {
490                         if (!message) {
491                                 snprintf(msg_buf, sizeof msg_buf,
492                                          "compression method %s not configured",
493                                          compress_name);
494                                 message = msg_buf;
495                         }
496                 } else
497                         error("junk in compressed archive");
498                 if (state != Reset)
499                         error("junk in compressed archive");
500                 this_header = saved_offset + my_inptr;
501                 buf += my_inptr;
502                 len -= my_inptr;
503         }
504         dir_utime();
505         kfree(name_buf);
506         kfree(symlink_buf);
507         kfree(header_buf);
508         return message;
509 }
510
511 static int __initdata do_retain_initrd;
512
513 static int __init retain_initrd_param(char *str)
514 {
515         if (*str)
516                 return 0;
517         do_retain_initrd = 1;
518         return 1;
519 }
520 __setup("retain_initrd", retain_initrd_param);
521
522 extern char __initramfs_start[];
523 extern unsigned long __initramfs_size;
524 #include <linux/initrd.h>
525 #include <linux/kexec.h>
526
527 static void __init free_initrd(void)
528 {
529 #ifdef CONFIG_KEXEC
530         unsigned long crashk_start = (unsigned long)__va(crashk_res.start);
531         unsigned long crashk_end   = (unsigned long)__va(crashk_res.end);
532 #endif
533         if (do_retain_initrd)
534                 goto skip;
535
536 #ifdef CONFIG_KEXEC
537         /*
538          * If the initrd region is overlapped with crashkernel reserved region,
539          * free only memory that is not part of crashkernel region.
540          */
541         if (initrd_start < crashk_end && initrd_end > crashk_start) {
542                 /*
543                  * Initialize initrd memory region since the kexec boot does
544                  * not do.
545                  */
546                 memset((void *)initrd_start, 0, initrd_end - initrd_start);
547                 if (initrd_start < crashk_start)
548                         free_initrd_mem(initrd_start, crashk_start);
549                 if (initrd_end > crashk_end)
550                         free_initrd_mem(crashk_end, initrd_end);
551         } else
552 #endif
553                 free_initrd_mem(initrd_start, initrd_end);
554 skip:
555         initrd_start = 0;
556         initrd_end = 0;
557 }
558
559 #ifdef CONFIG_BLK_DEV_RAM
560 #define BUF_SIZE 1024
561 static void __init clean_rootfs(void)
562 {
563         int fd;
564         void *buf;
565         struct linux_dirent64 *dirp;
566         int num;
567
568         fd = sys_open("/", O_RDONLY, 0);
569         WARN_ON(fd < 0);
570         if (fd < 0)
571                 return;
572         buf = kzalloc(BUF_SIZE, GFP_KERNEL);
573         WARN_ON(!buf);
574         if (!buf) {
575                 sys_close(fd);
576                 return;
577         }
578
579         dirp = buf;
580         num = sys_getdents64(fd, dirp, BUF_SIZE);
581         while (num > 0) {
582                 while (num > 0) {
583                         struct stat st;
584                         int ret;
585
586                         ret = sys_newlstat(dirp->d_name, &st);
587                         WARN_ON_ONCE(ret);
588                         if (!ret) {
589                                 if (S_ISDIR(st.st_mode))
590                                         sys_rmdir(dirp->d_name);
591                                 else
592                                         sys_unlink(dirp->d_name);
593                         }
594
595                         num -= dirp->d_reclen;
596                         dirp = (void *)dirp + dirp->d_reclen;
597                 }
598                 dirp = buf;
599                 memset(buf, 0, BUF_SIZE);
600                 num = sys_getdents64(fd, dirp, BUF_SIZE);
601         }
602
603         sys_close(fd);
604         kfree(buf);
605 }
606 #endif
607
608 static int __init populate_rootfs(void)
609 {
610         char *err = unpack_to_rootfs(__initramfs_start, __initramfs_size);
611         if (err)
612                 panic("%s", err); /* Failed to decompress INTERNAL initramfs */
613         if (initrd_start) {
614 #ifdef CONFIG_BLK_DEV_RAM
615                 int fd;
616                 printk(KERN_INFO "Trying to unpack rootfs image as initramfs...\n");
617                 err = unpack_to_rootfs((char *)initrd_start,
618                         initrd_end - initrd_start);
619                 if (!err) {
620                         free_initrd();
621                         goto done;
622                 } else {
623                         clean_rootfs();
624                         unpack_to_rootfs(__initramfs_start, __initramfs_size);
625                 }
626                 printk(KERN_INFO "rootfs image is not initramfs (%s)"
627                                 "; looks like an initrd\n", err);
628                 fd = sys_open("/initrd.image",
629                               O_WRONLY|O_CREAT, 0700);
630                 if (fd >= 0) {
631                         ssize_t written = xwrite(fd, (char *)initrd_start,
632                                                 initrd_end - initrd_start);
633
634                         if (written != initrd_end - initrd_start)
635                                 pr_err("/initrd.image: incomplete write (%zd != %ld)\n",
636                                        written, initrd_end - initrd_start);
637
638                         sys_close(fd);
639                         free_initrd();
640                 }
641         done:
642 #else
643                 printk(KERN_INFO "Unpacking initramfs...\n");
644                 err = unpack_to_rootfs((char *)initrd_start,
645                         initrd_end - initrd_start);
646                 if (err)
647                         printk(KERN_EMERG "Initramfs unpacking failed: %s\n", err);
648                 free_initrd();
649 #endif
650                 /*
651                  * Try loading default modules from initramfs.  This gives
652                  * us a chance to load before device_initcalls.
653                  */
654                 load_default_modules();
655         }
656         return 0;
657 }
658 rootfs_initcall(populate_rootfs);